These are no more "errors" than is the common practice in Trinity-style methodology of assuming that historical investors invested using investment vehicles that, in reality, would have been unavailable to them (like index funds). The purpose of these types of back-tests is not to determine how actual investors in the past would have actually fared subject to actual then-applicable real-world constraints; instead, it's to determine how hypothetical investors in the past would have fared (using actual historical market performance) if it were somehow possible for them to invest using the investment options available today. Of course, this does raise philosophical questions about the appropriateness (and usefulness) of the entire back-testing exercise -- see the discussion on that point in the handful of posts immediately above, and in the mortgage thread cited in reply # 53.
If I understood correctly, the graph attempts to depict not just market performance, but a comparison of "investing" marginal dollars into paying down a mortgage
instead of investing in the stock market.
While historical stock market return data has its own issues, as you mention, I'm not picking bones with historical stock market performance in isolation, but rather in this context where that data is graphed against the ten-year Treasury yield as a proxy for historical mortgage data.
The graph depicts, in essence, stock returns versus bond returns, which isn't what it purports to depict.
In fact, I would argue that the multiple options currently available in the mortgage market, combined with the changes in mortgage loans seen in the 20th century, makes the attempt to reduce the comparison of investing the marginal dollar (in the market versus a mortgage) impossible to graph reasonably.
While US mortgage interest rates move in tandem with the 10-year Treasury, 15-year and 30-year fixed-rate mortgages nearly always sport significantly higher interest rates because of the length of the rate lock. Because of this, the 30-year Treasury note would make a better proxy for the 30-year fixed-rate mortgage. Moreover, as I mentioned, 15- and 30-year conforming loans didn't exist before the 1950s. Therefore using the 10-year as a proxy doesn't work well at all. I suppose you could argue that it's better than nothing, but then I would say you shouldn't put much faith in such a poor model.
Note, by the way, that I'm arguing against my own actions (I carry a mortgage when I could easily pay it off), because the 10-year Treasury is easier to beat that the higher 30-year fixed-rate mortgage I'm paying.